Dis cus si on Paper No. 12-054
An Integrated Assessment Model with Endogenous Growth
Michael Hübler, Lavinia Baumstark, Marian Leimbach, Ottmar Edenhofer, and Nico Bauer
Dis cus si on Paper No. 12-054
An Integrated Assessment Model with Endogenous Growth
Michael Hübler, Lavinia Baumstark, Marian Leimbach, Ottmar Edenhofer, and Nico Bauer
Download this ZEW Discussion Paper from our ftp server:
http://ftp.zew.de/pub/zew-docs/dp/dp12054.pdf
Die Dis cus si on Pape rs die nen einer mög lichst schnel len Ver brei tung von neue ren For schungs arbei ten des ZEW. Die Bei trä ge lie gen in allei ni ger Ver ant wor tung
der Auto ren und stel len nicht not wen di ger wei se die Mei nung des ZEW dar.
Dis cus si on Papers are inten ded to make results of ZEW research prompt ly avai la ble to other eco no mists in order to encou ra ge dis cus si on and sug gesti ons for revi si ons. The aut hors are sole ly
respon si ble for the con tents which do not neces sa ri ly repre sent the opi ni on of the ZEW.
Non-technical Summary
We introduce endogenous directed technical change into integrated multi-region pol-
icy assessment based on macroeconomic theory and evidence. Technical progress can
be directed towards labor or energy savings. We distinguish expenditures on innovation
and imitation, in other words international technology transfer, and take the role of cap-
ital investment in implementing new technologies into account. We study the regional
pattern of mitigation costs in form of consumption losses induced by a global carbon
budget-based climate policy. Mitigation costs turn out to be robust with respect to a
variation in the parameter values within our model of endogenous growth – except the
effectivity (in terms of technical progress per unit of investment) of energy specific rela-
tive to labor specific innovation and imitation expenditures. This result suggests that an
increased effectivity of energy specific innovation and imitation expenditures can over-
compensate rising emissions due to economic growth and thus reduce mitigation costs.
However, the effectivity is exogenously determined by technological restrictions and at
best partly susceptible by economic policy in the short-run. In the medium- to long-run,
the effectivity could be improved by emphasizing energy efficiency aspects in education,
basic research, infrastructure, structural change and capacity building for R&D and for
the absorption of foreign technologies.
Motivated by recent announcements of providing financial and technological transfers
for developing countries, we examine interregional energy saving technology transfer
starting at different points of time. Herein, all regions gain from technology transfer
due to its growth effect. China appears as the main beneficiary of early technology
transfer because of strong international technology spillovers, followed by the region of
the developing countries.
Our results suggest that endogenous energy specific technical progress has already
been strongly exploited in the business as usual baseline and created baseline consump-
tion gains. This might be due to perfect foresight and regionally weighted global op-
timization in combination with energy scarcity. But in reality, the baseline gains are
not fully exploited on a market base due to risks and frictions and missing information.
Realizing these gains probably requires policy support in form of an integrated climate
and energy policy concept. Herein, early energy saving technology transfer could be a
’carrot’ to encourage developing countries to engage in climate protection.
Das Wichtigste in Kurze
In diesem Artikel, fuhren wir endogenen gerichteten technischen Fortschritt
basierend auf makrookonomischer Theorie und Empirie in die integrierte multiregionale
Politikanalyse ein. Technischer Fortschritt kann zur Einsparung der Inputfaktoren Ar-
beit und Energie verwendet werden. Wir unterscheiden zwischen Ausgaben fur Innova-
tion und Imitation (internationalen Technologietransfer) und berucksichtigen die Bedeu-
tung von Kapitalinvestitionen fur die Implementierung neuer Technologien. Wir unter-
suchen die regionale Verteilung der Vermeidungskosten, die sich aus einer Klimapolitik
mit globalem CO2-Budget ergeben. Die Vermeidungskosten erscheinen weitgehend ro-
bust im Hinblick auf die Variation der Parameterwerte im Modell des endogenen Wach-
stums. Es zeigt sich jedoch, dass eine hohere Effektivitat (in Form von technischem
Fortschritt pro Investitionseinheit) von energiespezifischer relativ zu arbeitsspezifischer
Innovation und Imitation wachstumsbedingt steigende Vermeidungskosten uberkompen-
sieren kann. Diese Effektivitat ist jedoch exogen durch technische Restriktionen gegeben
und bestenfalls teilweise kurzfristig durch Wirtschaftspolitik beeinflussbar. Mittel- bis
langfristig konnte die Effektivitat durch die Betonung von Energieeffizienzaspekten in
Bildung, Grundlagenforschung und Infrastruktur sowie durch die Verbesserung der Kom-
petenz in Forschung und Entwicklung und in der Absorption auslandischer Technologien
verbessert werden.
Motiviert durch die Ankundigungen von finanziellen und technischen Transfers
an Entwicklungslander, untersuchen wir interregionalen energiesparenden Technologi-
etransfer, der zu unterschiedlichen Zeitpunkten beginnt. Dabei profitieren alle Regionen
von dem Wachstumseffekt des Technologietransfers. China erscheint als Hauptprofiteur
des fruhzeitigen Technologietransfers bedingt durch starke internationale Technologie-
Spillovers, gefolgt von der Modellregion der Enwicklungslander.
Unsere Ergebnisse zeigen, dass der energiespezifische technische Fortschritt bereits
im Basislauf (BAU) weitgehend ausgeschopft wird. Dies kann an perfekter Voraussicht
und an regional gewichteter globaler Optimierung in Kombination mit Energieknappheit
im Basislauf liegen. In der Realitat wird dieses Potenzial wegen Risiken, Markthemnissen
und mangelnder Information nicht voll ausgeschopft, so dass ein integriertes Klima- und
Energiekonzept sinnvoll erscheint. Dabei kann fruhzeitiger Technologietransfer einen
Anreiz fur Entwicklungslander bieten, sich an einem Klimaabkommen zu beteiligen.
An Integrated Assessment Model with Endogenous Growth
Michael Hubler∗, Lavinia Baumstark,
Marian Leimbach, Ottmar Edenhofer, Nico Bauer
August 22, 2012
Abstract
We introduce endogenous directed technical change into numerical integrated
climate and development policy assessment. We distinguish expenditures on
innovation (R&D) and imitation (international technology spillovers) and consider
the role of capital investment in creating and implementing new technologies. Our
main contribution is to calibrate and numerically solve the model and to examine
the model’s sensitivity. As an application, we assess a carbon budget-based climate
policy and vary the beginning of energy-saving technology transfer. Accordingly,
China is a main beneficiary of early technology transfer. Herein, our results
highlight the importance of timely international technology transfer for efficiently
meeting global emission targets. Most of the consumption gains from endogenous
growth are captured in the baseline. Moreover, mitigation costs turn out to be
insensitive to changes in most of the parameters of endogenous growth. A higher
effectivity of energy-specific relative to labor-specific expenditures on innovation
and imitation reduces mitigation costs, though.
JEL Classifications: O11, O30, O44, O47, Q32
Keywords: endogenous growth, directed technical change, technology transfer,
integrated assessment, carbon budget, China
∗Corresponding, leading author, email: [email protected], tel: +49-621-1235-340, fax: +49-621-1235-4340, Centre for European Economic Research, Postfach 103443, 68034 Mannheim, Germany. All au-thors’ affiliation: Potsdam Institute for Climate Impact Research, Telegraphenberg A31, 14412 Potsdam,Germany, email: [email protected], [email protected], [email protected],[email protected]. We thank Elmar Kriegler for his help regarding IPCC scenarios. We thankthe Leibniz Association (WGL) for financial support.
1 Introduction
Innovation as well as imitation and international diffusion of technologies can be a
key for successfully coping with poverty and climate change. Herein, (climate) policy
interventions have an impact on the strength and direction of innovation, imitation and
technology diffusion. Therefore, a (climate) policy analysis that takes these aspects into
account requires a rigorous model of endogenous directed technical progress. Setting up
such a model and calibrating it to real world data is the first and main contribution of
this paper. Due to the uncertainties in the parameter values in a model of endogenous
growth, we conduct a careful sensitivity analysis. This is the second contribution of this
paper.
It is widely agreed that the OECD countries bear the main responsibility for climate
change while the developing countries will bear most of its impacts. Private investment
on a national or international scale is expected to bring about the relevant capacities
and technologies for climate change mitigation and adaptation. China as a prominent
example has successfully improved its energy productivity and has become a leading
producer and exporter of clean energy equipment. But in general, many developing
and emerging economies lack in financial resources, knowledge, technological capabili-
ties and the ability to adopt foreign technologies. International trade policy and patent
regulation (WTO and TRIPS1) can on the one hand spur innovation but on the other
hand hinder international technology diffusion and technological catching up. There-
fore, many economies will probably not be able to achieve technical progress, economic
development and carbon emissions reductions simultaneously within a short time frame.
Thus international support will be required.
Therefore, in recent climate negotiations (Bali Roadmap 2007, Copenhagen 2009 and
Cancun 2010 summit), developing countries called for financial and technological sup-
port for mitigation, and industrialized countries announced to provide such support. So
far, the Kyoto Protocol has enabled international financing in (and technology transfer
to) developing countries within the Clean Development Mechanism (CDM) framework.
Herein, China has been the biggest seller of CDM credits with a market share of 72%
in 2009 (Kossoy and Ambrosi 2010, section 4). The total volume of CDM transac-
tions amounted to US-$ 6.5 billion in 2008 and only US-$ 2.7 billion in 2009 (Kossoy
and Ambrosi 2010, section 4). Moreover, developing countries can receive such support
through technology funds like the World Bank Climate Investment Funds (World Bank
2010) as announced at the Cancun 2010 summit. In particular, industrialized countries
1World Trade Organization and Trade-Related Aspects of Intellectual Property Rights.
1
announced future transfers amounting to US-$ 100 billion per annum by 2020 in the
Copenhagen Accord. This volume exceeds the volumes of annual financial transfers
within the CDM framework cited above by far. However, no legally binding commit-
ments have been achieved that settle which countries will pay and receive how much
beginning at which date. This uncertainty gives rise to the question how mitigation
costs of different regions are affected by postponing international technology transfer.
Against this background, in this article, we apply our model of endogenous growth to
the assessment of mitigation costs induced by a carbon budget-based policy and the
costs of delaying international technology transfer.2 Thereby, it intends to contribute
to the literature that discusses the future of the Kyoto Protocol against the backdrop
of efficient (carbon) markets and North-South equity (c.f. Chichilnisky and Heal 2000;
Chichilnisky and Sheeran 2009). This is the third contribution of this paper.
Our model approach refers to state-of-the-art theoretical models of endogenous
growth.3 Product variety models in the style of Romer (1990) describe growth as a
process that stems from an increasing number of innovative intermediate products (e.g.
Grossman and Helpman 1991). Product quality models in the style of Aghion and
Howitt (1992) rather describe growth as a process that stems from quality improvements
of products wherein new varieties replace old varieties, which is also called ’creative de-
struction’. We refer to the latter model type, however on a stylized macro level without
treating profit maximizing firms explicitly into account. Acemoglu et al. (2003, 2006)
provide microfoundations and a rigorous analysis of the influence of the distance between
the technology in practice and the technology frontier (along the lines of the seminal
contribution by Nelson and Phelps 1966). They show that an imitation-based strategy
is preferable when being further away from the technology frontier while an innovation-
based strategy is preferable when being closer to the technology frontier. We follow this
idea by including a ’distance to technology frontier’ term (more specifically a ’technol-
ogy pool’ term) in our model. Herein, the model allows an endogenous simultaneous
choice between innovation and imitation which are treated as substitutes. It basically
reproduces the findings by Acemoglu et al. (2003, 2006) endogenously. Furthermore,
we follow approaches in the style of Arrow (1962) such as Greiner and Semmler (2002)
that view learning related to capital investment as a driver of technical progress. In our
context, the positive impact of capital investment on technical progress in an economy
is a supplement to the following consideration: New technologies such as energy-saving
2We leave the specific channels – such as FDI – and policy instruments – for instance a technologyfund – for achieving international technology transfer open.
3As comprehensively described by Aghion and Howitt (2009), chapter 4 and Acemoglu (2009), chap-ters 14, 15 and 18.
2
technologies that exist as blueprints become increasingly used in the economy through
capital investment. As a result, they become increasingly embodied in the new capital
stock and raise its productivity. We implement this feature in the style of the Schum-
peterian model as a novel theoretical detail. Finally, we follow the literature in the style
of Acemoglu (2002) that emphasizes the possibility to direct technical change towards
specific factors depending on the abundance of factors or relative factor prices. Tech-
nical progress directed towards a certain factor will reduce the demand for this factor
(factor-saving technical progress) when the elasticity of substitution between the pro-
duction factors is smaller than one, which is the case in our model (in the upper CES
level).
However, endogenous growth along these lines of the theoretical literature has not yet
been fully worked out in an integrated assessment framework. Therefore, it is our main
contribution to implement endogenous, directed technical progress resulting in fully en-
dogenous economic growth into our multi-region integrated assessment model. Therein,
our approach contributes to the literature that numerically describes endogenous innova-
tion (e.g. Goulder and Schneider 1999, Popp 2004, 2006, Edenhofer et al. 2005, Kemfert
2005, Otto et al. 2007, 2008, Loschel and Otto, 2009, Gerlagh 2008) and international
technology spillovers (e.g. Diao et al. 2005, Leimbach and Baumstark 2010, Hubler
2011). Our model is mostly comparable to the integrated assessment model WITCH
(Bosetti et al. 2006). The original version of WITCH, described and applied by Bosetti
et al. (2008), focuses on disembodied international technology spillovers of energy-
specific R&D (research and development). Herein, the strength of spillovers depending
on the distance to the technology frontier has an inverted U shape. This means, technol-
ogy spillovers are highest at a medium distance to the technology frontier. Bosetti et al.
(2011) apply WITCH to show that innovation policy in combination with climate policy
results in substantial efficiency gains. Our model model additionally allows for R&D
and international spillovers that are directed to labor inputs and assumes that spillovers
increase in the distance to the technology frontier. The modified version, used by Nicita
et al. (2009), also allows for the direction of R&D towards energy or non-energy (cap-
ital and labor) inputs and thus endogenizes crowding out effects. Their model version
seems not to model international technology spillovers, though. Their analysis shows
that climate policy shifts R&D more towards energy inputs while R&D declines in total
since total output declines. Our model combines both effects, international spillovers
and directed technical change, with respect to labor and energy productivity. Com-
pared to WITCH, our model also represents international (and ’intertemporal’) trade
in goods which aggravates the numerical solution of the multi-region model. Moreover,
3
compared to WITCH, our model represents endogenous resource extraction that yields
a kind of Hotelling path. On the contrary, we do not take climate damages into ac-
count in our model. We do not model international R&D spillovers of energy conversion
technologies either but apply a global learning curve for the technologies wind and solar
photovoltaic. This means, domestic investment costs of these technologies decrease in
the installed capacity world-wide. This clearly has an impact on technology choice and
technology diffusion. We apply our model to an analysis that is new in the literature:
The effects of delaying international technology diffusion.
In our policy analysis, our model of endogenous growth will be embedded into the
integrated assessment model REMIND (Refined Model of Investment and Technological
Development, Leimbach et al. 2010a, c.f. Figure 3 and Figure 4 in the Appendix),
a Ramsey type model of intertemporally optimal investment in physical capital and
energy technology capacities. The model version under scrutiny consists of five world
regions and includes trade (in a composite commodity, coal, gas, oil, uranium and carbon
emissions permits) between these regions. Technology spillovers are controlled in a cen-
tralized way. International trade is subject to an intertemporal trade budget restriction
following Negishi (1972) which creates a decentralized solution for trade. The macro
model is coupled with an energy system module that represents several energy sources
and related capacities of energy technologies (coal, gas, oil, uranium, hydro, biomass,
solar, wind, geothermal, carbon capture and storage (CCS) of coal, gas and biomass; c.f.
Leimbach et al. 2010a, b). The energy system module includes endogenous investment
into capacities of different energy technologies as well as learning-by-doing of wind and
solar technologies following the literature that emphasizes learning effects (e.g. Crassous
et al. 2006, Kahouli-Brahmi 2008). The energy system module takes increasing costs
of resource extraction into account as well as operation and maintenance costs. Carbon
emissions stemming from fossil fuels burned in production and consumption processes
can be translated into resulting temperature increases in a climate module (Tanaka and
Kriegler 2007). The climate module is not used in this analysis of endogenous growth,
though. The time horizon under scrutiny is 2005 until 2100 in five-year steps.
Section 2 derives our model of endogenous growth from economic theory. Section
3 describes the numerical calibration and shows baseline simulation results. Section
4 applies the model to the assessment of a carbon budget-based climate policy and
the delay of energy-saving technology diffusion within the integrated assessment model
REMIND (Leimbach et al. 2010a). Section 5 critically discusses the model results
and carries out a sensitivity analysis. Section 6 concludes by deriving policy implications.
4
2 Model
We derive the implementation of directed technical change in an intertemporal optimiza-
tion framework in four steps from economic theory: (1) We derive the effect of R&D
expenditures on the progress of innovation from a Schumpeterian model of growth. (2)
We take investment in physical capital as a driver of innovation into account. (3) We
implement interregional technology spillovers. (4) We allow for the direction of tech-
nical change towards labor or energy. (Table 6 and Table 7 in the Appendix provide
an overview of the sets, variables and parameters of our model of endogenous growth.)
In the following, we will describe the first three drivers of technical change separately
and finally add them up including the fourth effect. Herein, our model represents the
macroeconomic level without modeling firms explicitly. Our treatment of energy-specific
technical change refers to productivity gains in using energy in final production. We do
not model international R&D spillovers of energy conversion technologies but apply a
global learning curve for new technologies like wind and solar photovoltaic. This means,
domestic investment costs of these technologies decrease with the installed capacity
world-wide. Table 6 in the Appendix summarizes the model notation.
2.1 R&D expenditures
With respect to modeling endogenous growth, we follow the Schumpeterian view of
quality improvements as a driver of economic growth-based on the description by Aghion
and Howitt (2009), chapter 4. We start with a one-sector production function Y which
is increasing in technology A. Both are macroeconomic aggregates so that A =∫ 10 Ajdj
can be interpreted as an average of individual productivities of firms or sectors j in the
economy. In each period a firm spends Rj on R&D. Each firm is able to keep part of
the generated knowledge as firm-specific knowledge so that it has some monopolistic
power and earns a profit. In other words, each firm holds a patent. The same intuition
applies to non-profit research institutions in form of earning non-monetary profits such as
publications, reputation and political influence, so that we may also interpret non-profit
organizations as firms. Now we aggregate individual expenditures to macroeconomic
expenditures R =∫ 10 Rjdj. On the macro level, R may also include public spending on
education, basic research, infrastructure etc., which enhances invention and innovation
in the economy. Each innovation process is due to uncertainty regarding its outcome.
By the law of large numbers, expenditures will lead to a successful innovation with
probability µ and will not lead to a successful innovation with probability 1− µ on the
macro level. Herein, µ is increasing in R which is endogenously determined. Innovation
processes are not only due to uncertainties but also inefficiencies, for example because
5
of short time horizons that managers consider for the innovation investment decisions
or insufficient protection of intellectual property rights. Such inefficiencies of innovation
markets are implicitly captured by the innovation probability µ. Higher inefficiencies
reduce the probability of success when spending a certain volume for R&D resulting in
a lower µ. More specifically, following Aghion and Howitt (2009), chapter 4, we write:4
µ = λR
(Rt
ARt
)σR
(1)
t ∈ {2005, 2010, ..2100} denotes time, more specifically years increasing in five-year steps.
Rt denotes aggregated macroeconomic R&D expenditures at time t, and ARt stands for
the aggregated individual productivities at time t regarding R&D expenditures R. λR
determines the impact of R&D expenditures on the probability of success in a linear
fashion. σR creates a decreasing marginal effect of R&D expenditures on the probability
of success with rising expenditures, where 0 < σR < 1. Assuming that a new technology
is γ > 1 times as productive as the previous technology, the rate of innovation-based
technical progress gR can be derived in the following way:
ARt+5 = µγAR
t + (1− µ)ARt ⇔
ARt+5 −AR
t
ARt
= µ(γ − 1) =: gR (2)
In case of a successful innovation, the new technology γARt will be applied. In case of no
success, the old technology ARt will be used further.5 However, the implementation in
the REMIND model does not take profit maximization of firms and monopolistic power
due to successful innovations explicitly into account. Thus, ARt+5 represents an aggregate
or average technology level of the economy in period t. On this aggregate level, ARt+5
can be treated as deterministic representing a certain share µ of successful innovators
in all firms. Our model simulations will treat the technology level as a deterministic
variable too.
2.2 Investment in physical capital
Additionally, there is an interaction of investment in knowledge creation and investment
in capital. On the one hand, the literature based on Arrow (1962) sees knowledge
as a by-product of capital accumulation. On the other hand, viewing knowledge as a
public good, innovations need time for diffusion through the economy, and they require
investment in capital in order to be implemented into production facilities. Therefore,
4We will add region- and factor-specific indexes in the final set of equations.5One may add depreciation of knowledge which is less common in theoretical growth models than in
applied assessment models.
6
we extend the Schumpeterian point of view in a novel setting in the following general
form:
ARIt+5 = (1 + gR)A
RIt
(It
Kt+5
)σI
+ARIt
[1−
(It
Kt+5
)σI]
(3)
ARIt is the productivity correlated with R&D expenditures and investments into physical
capital. It is investment in capital, and Kt+5 = (1− δ)Kt + It is the new capital stock,
where δ is the depreciation rate. We assume σI = 1.6 Then, according to the equation
above, the fraction of the capital stock remaining after depreciation that is renewed
by investment uses the newest technology (1 + gR)ARIt . The remaining fraction of the
capital stock still uses the old technology ARIt . As a consequence, the implementation of
existing new technologies in production depends on investment, as observed in reality.
We can now simplify the equation above and replace gR:
ARIt+5 = ARI
t
[1 + (γ − 1)λR
(Rt
ARIt
)σR(
ItKt+5
)σI]
(4)
2.3 International technology spillovers
In the next step, we will add international technology diffusion following the same line
of argumentation and the same specification as before. There are basically two differ-
ences to the previous specification. We now assume that expenditures S encompass
expenditures on fostering international technology diffusion instead of innovation. They
include expenditures of firms on the imitation and adoption of foreign technologies as
well as publicly funded projects that enhance the diffusion of technologies. Besides
this re-interpretation, a new technology still appears with probability µ as described by
equation (1), but now we assume that each productivity increase, previously occurring
at the rate γ − 1, occurs endogenously. This productivity increase depends inversely on
the technology level of the recipient economy relative to the world technology pool A as
suggested by Acemoglu (2009), chapter 18 (c.f. Griffith et al. 2003 reconciling theory
and evidence). The rate of technical progress now reads:
gS := λS
(St
ASIt
)σS(ASI
t
ASIt
)σA
(5)
In general, it is possible that λR = λS and that σR = σS since innovation and imita-
tion or diffusion are driven by different processes. The term(AtAt
)σA
implies that the
larger an economy’s technology gap relative to the world technology pool the higher is
6Allowing 0 < σI < 1 means, it becomes increasingly difficult or costly to replace a larger fraction ofthe capital stock by the newest technology.
7
its growth rate.7 As a theoretical result, all economies will grow at the same rate but
at different relative distances to the technology pool level depending on their absorptive
capacities in the long-run steady state. As suggested by Acemoglu (2009), we compute
the world technology pool as the arithmetic average of the technology levels of all re-
gions. As a consequence, all regions contribute to the world technology pool. In some
’technology frontier’ specifications often used in the literature, on the contrary, only the
technology leader pushes the frontier forward and thus contributes to the global stock of
technological knowledge. Herein, we implicitly assume that technological knowledge is
heterogenous so that the best available technology does not incorporate all technological
know-how, but instead all inventors contribute to a common knowledge pool. Taking
again the role of investment into account yields:
ASIt+5 = ASI
t
[1 + λS
(St
ASIt
)σS(
ItKt+5
)σI(ASI
t
ASIt
)σA]
(6)
2.4 Directed technical change
Following Acemoglu (2002), we now take directed, i.e. factor-specific technical progress
into account. In each region the REMIND production function includes the input factors
capital, K, labor, L, and energy, E, and has the following form:
Yt =
[αK(AKKt)
σY −1
σY + αL(ALtLt)σY −1
σY + αE(AEtEt)σY −1
σY
] σYσY −1
(7)
While AK is kept constant, ALt and AEt rise endogenously representing labor- and
energy-specific technical progress. Each type of endogenous technical progress is mod-
eled as described above. We choose the elasticity of substitution 0 < σY = 0.5 < 1 so
that the production factors are gross complements. In this case, according to Acemoglu
(2002), energy augmenting technical progress, i.e. growth of AEt, is labor biased, i.e. it
creates excess demand for labor rather than for energy and raises the marginal product
of labor more than the marginal product of energy. This means, energy scarcity, e.g.
due to climate policy, induces energy saving (energy augmenting) technical progress.
After combining the effects (1) to (4), i.e. adding up Equations (4) and (6), At+5 =
ARIt+5 +ASI
t+5, and introducing a factor index i = {L,E} and a region index r, we obtain
7According to Acemoglu (2009), chapter 18 one may set σA ≥ 1 so that economies farther away fromthe technology pool level have a stronger advantage with respect to technology diffusion.
8
the final equation below:
Arit+5 = Arit
{1 + λiλrt
[λR
(Rrit
Arit
)σR
+ λS
(Srit
Arit
)σS(
Ait
Arit
)σA](
IrtKrt+5
)σI}
(8)
Since Equations (4) and (6) are combined in an additive way, innovation and imitation
are treated as substitutes. As a result and in accordance with the micro-foundations
described by Acemoglu et al. (2003, 2006), imitation is more beneficial farther away from
the technology frontier (in our case technology pool), while innovation (Equation 4) is
more beneficial closer to the technology frontier. Moreover, we extend the parameters
λR and λS that determine the strength of innovation and imitation by a factor-specific
differential λi and an interregional differential λrt. Herein, λi might differ between
energy and labor due to technological reasons, i.e. the value of energy saved by a certain
volume of R&D investment can differ from the value of labor saved by the same volume
of R&D investment.8 λrt is – besides other aspects – determined by the educational level
(human capital) of the respective region. The important role of education for innovation
and imitation (absorptive capacity) has often been emphasized in the theoretical and
empirical literature (Nelson and Phelps 1966, Benhabib and Spiegel 2005, Kneller 2005).
Herein, regional education levels may change over time and in particular converge to
equal levels across regions in the distant future.
The overall objective of our Ramsey type optimization model is the weighted sum of
utility drawn from per capita consumption across all regions, cumulated and discounted
over the time horizon. Regional weighting follows Negishi (1972). Expenditures re-
lated to knowledge creation, which we call Rrit and Srit create costs in form of foregone
consumption Crt like usual investment in capital Irt. In other words, final output can
directly be used as an intermediate input for the creation of knowledge so that consump-
tion in each region is given by:
Crt = Yrt +Mrt − Irt −Qrt −RrLt −RrEt − SrLt − SrEt (9)
The marginal product of physical capital Krt rises as a consequence of technical
progress which stimulates capital investment Irt. Additionally, the REMIND model
encompasses an energy system module that distinguishes several energy sources (coal,
gas, oil, uranium, hydro, biomass, solar, wind, geothermal). Investments into capacities
of the related energy technologies, fuel, operation and maintenance costs are also
subtracted from the budget like investment in physical capital as a production factor.
8For example, a state-of-the-art washing machine will save energy and save time spent for operatingit to different extents.
9
They are subsumed under energy system expenditures Qrt in the equation above.
Finally, Mrt denotes net imports of each region. The REMIND model applies an
algorithm based on Negishi (1972) that adjusts regional weights in the welfare function
such that the intertemporal trade budget is equal to zero for all regions in 2150. This
algorithm creates a decentralized Nash solution with respect to interregional trade. Mrt
represents trade in goods and services as well as international transfers. The latter will
be important in our analysis of financing international technology diffusion. Obviously,
any interregional transfer Mrt can be used for consumption and the various kinds of
investment in physical capital, energy technology capacities, innovation and imitation
such that the resulting investment pattern is optimal.
2.5 Energy system
The energy system9 is mainly characterized by a rich technology portfolio and diversity
of primary energy sources, namely coal, oil, natural gas, uranium, biomass, solar, wind,
hydro and geothermal. (Their developments over time are shown in Figure 5e in the
Appendix.) Secondary energy carriers are electricity, heat, hydrogen, other liquids, solid
fuels, gases, transport fuel petrol and transport fuel diesel. Therein, the model assumes
a linear substitution possibility between competing technologies for secondary energy
production. From the energy use side, total energy used in production and consumption
consists of stationary energy and transport energy (see Figure 4 in the Appendix).
Stationary energy is further split into electricity and non-electricity energy. Transport
energy is further split into liquid fuels and hydrogen.
System integration costs and resource availability (resource quality) lead to marginal
costs of energy generation that rise in the annual technology deployment and in resource
extraction respectively. They therefore dampen the expansion of the energy technologies
represented in the model. Therein, the availability of (exhaustible) fossil resources is
modeled via supply functions that include cumulative extraction costs and take regional
resource endowments into account.
Endogenous technological learning, on the contrary, lowers marginal costs of wind
and solar technologies depending on their cumulative technology deployment (cumulated
energy generation) following Barreto (2001). It creates a kind of positive feedback and
therefore helps expand wind and solar capacities over time. This can partly explain
the expansion of wind and in particular solar power in the model (see Figure 5e in the
9For further details see http://www.pik-potsdam.de/research/sustainable-solutions/models/
remind/remind-code.
10
Appendix). Nonetheless, biomass expands even more over time although it does not
include a learning mechanism. Non-renewable energies are not due to learning effects.
A larger endowment with a fossil resource, say gas, results in lower marginal extraction
and energy generation costs and thus in the expansion of investments into its capacity.
As a consequence of carbon pricing, fossil energies are clearly reduced over time under
climate policy (see Figure 5e). Allowing for learning effects, say in gas power, could
enhance fossil fuel use under climate policy to some extent. Therein, we do not allow for
endogenous R&D in any energy technologies. R&D in energy technologies could help
expand non-fossil technologies and improve fossil technologies.
The macroeconomic module and the energy system module are integrated using a
hard-link that ensures simultaneous equilibria on all energy and capital markets (c.f.
Bauer et al. 2006). From the macroeconomic perspective, this means that the in-
vestments in energy conversion technologies and the demand for final energy carriers
required for the generation of economic income are in balance. From the energy sys-
tem perspective, the supply of primary energy carriers, the investments into energy
conversion technologies and thus the supply of final energy carriers are also in balance.
In equilibrium, the energy sector allocates the scarce resources of capital and pri-
mary energy to energy technologies in accordance with the prices for these factors, the
price for carbon emissions and the prices for final energy carriers that are supplied to
the macro module. The conversion technologies require investments into their capacity
stocks. These capacities limit the specific flows of energy converted from one form into
another form. For example, built gas power plant capacities limit the conversion of
gas into electricity. The technology richness and the diversity of energy technologies
imply that the equilibrium solution for technology choice regarding scale, timing and
location of each technology depends on the prices for capital and energy. Cumulative
consumption of exhaustible energy carriers and carbon emissions as well as changing
demand structures for final energy imply sequences of technology transitions. For ex-
ample, the transition from coal power to wind and finally solar power will be accelerated
if the carbon price increases and exceeds a certain threshold so that the technologies
become profitable. Investments in energy technologies are financed via the savings of
the representative consumer who in turn receives the return on investment.
3 Calibration
We aggregate the integrated assessment model REMIND to five world regions: INA con-
sists of Africa, Latin America, India and other Asia, i.e. it includes major low-income
countries. China is denoted by CHN. ROW consists of Middle East, Japan, Russia and
11
the rest of the world, i.e. it is on average a middle-income region. EUR consists of the
European Union EU 27. USA denotes the United States of America. The overall calibra-
tion follows the version REMIND-R ’RECIPE’ (Edenhofer et al. 2009). With respect to
the population scenario, we refer to the UN (2008) projections. The respective medium
scenario peaks at around 9 billion people at the end of the century. Within our model of
endogenous growth, we need to calibrate the parameters λi (factor specialty of technical
progress), λr2005 (education level in 2005), λR (coefficient of innovation expenditures),
λS (coefficient of imitation expenditures), σR (exponent of innovation expenditures), σS
(exponent of imitation expenditures), σA (exponent of the technology gap term) and σI
(exponent of the investment term). Table 7 in the Appendix summarizes the parameters
and the corresponding values that will be derived in the following. Our calibration strat-
egy is based on (1) econometrically estimated values, (2) historical statistical reference
values and (3) future reference values derived from existing scenario simulations:
3.1 Econometric estimations
Griffith et al. (2003) reconcile the theoretical literature on Schumpeterian endogenous
growth with the econometric literature on R&D, growth and convergence. They review
the empirical findings on the influence of R&D expenditures per GDP on productivity
growth as a macroeconomic social benefit and list some examples: Griliches and Licht-
enberg (1984) find values of 0.21–0.76, Schankerman (1981) finds 0.24–0.73, and Scherer
(1982, 1984) obtains 0.29–0.43. In general, this literature strand finds a positive and sta-
tistically significant influence of R&D expenditures on productivity growth. These values
translate into the R&D coefficient λR in our model. However, the findings differ across
studies depending upon the underlying data sample, the definition of R&D (private,
public or both) and the inclusion or exclusion of international R&D spillovers (Griffith
et al. 2003). Griffith et al. (2004) find values around 0.4 depending upon the model
specification (including R&D expenditures per GDP as a lagged variable). Zachariadis
(2003) also finds values around 0.4. In accordance with this literature strand, we set λR
= 0.4. Note that different to the econometric literature we include R&D expenditures
divided by the current technology level as in the theoretical literature (instead of R&D
expenditures divided by GDP) and additionally the share of capital investment in GDP.
Griffith et al. (2004) additionally include R&D expenditures per GDP multiplied
by the technology gap which corresponds to the term(
StAt
)σS(AtAt
)σA
in our model.
They find coefficients of 0.6–1.2. These coefficients translate into the R&D coefficient
λS for technology diffusion and imitation. Herein, different to our specification, Griffith
et al. (2004) include the technology gap term in logarithmic form, and they use the
12
technology frontier, i.e. the best available technology, instead of the average technology
level. Since our specification deviates from this econometric specification, we set λS to
a lower number of 0.12, which yields realistic productivity growth rates as described
below.
Furthermore, Zachariadis (2003) regresses the logarithmic rate of patenting in an
industry on the logarithmic R&D intensity based on a Schumpeterian model of growth.
This helps us set the exponent of R&D expenditures denoted by σR. Zachariadis (2003)
finds values around 0.2 for own-industry R&D (and about 0.6 for aggregate R&D). We
set σR = 0.1 in order to better match the historical R&D expenditures as described
below.
3.2 Historical data
The theoretical and econometric literature views education (human capital) as an im-
portant determinant of productivity growth through R&D and technology diffusion (c.f.
Nelson and Phelps 1966, Crespo et al. 2004, Benhabib and Spiegel 2005, Kneller 2005).
Herein, the absorptive capacity for the adoption of newly arriving technologies is sup-
posed to increase not only in education and skills but also in the existing infrastructure,
especially with respect to access to sources of knowledge and information technologies.
Also, the existing technologies in practice are supposed to ease the adoption of new
technologies. We follow this view by setting the coefficient λr2005 that affects both, in-
novation and diffusion of technologies, depending on region-specific levels of education
and infrastructure as a determinant of the absorptive capacity. We choose the param-
eters based on education and infrastructure indicators as reported by WDI (2010).10
Moreover, we assume that regions that lack in education and infrastructure catch up
over time so that λrt converges. Herein, we assume that all regions will reach the max-
imal value of one in 2100 (as illustrated in Figure 5 in the Appendix).
Table 1 confronts the results of our simulations for business as usual without climate
policy, BAU, with the reference data, REF, obtained from WDI (2010) and IEA (2010).
Herein, we compute averages over the time span 1996–2006 (in order to avoid the use of
outlier values). Obviously, the model results match the reference data well in many cases,
but there are also significant deviations, e.g. the growth rates of energy productivity in
10We examine primary, secondary and tertiary education enrolment and completion ratios as well asinfrastructure indicators such as internet and telephone access ratios. We set the highest value to oneand measure the other values relative to one. Then we compute the average of the rankings according tothe different indexes. The data in general yield the ranking USA, EUR, ROW, CHN, INA. We follow thisranking. However, it is difficult to make a decisive choice on the indicators to be included. Therefore,we adjust the education indicators such that the resulting GDP growth rates better match the historicaldata. This adjustment may also consider region size effects such that the regional aggregation chosendoes not arbitrarily influence the regional innovative performances.
13
Symbol Explanation Scen. INA CHN ROW EUR USA
g(Yr2005) GDP growth BAU: 4.7 10.1 3.5 2.8 3.5
REF: 4.0 9.2 2.4 2.5 3.0
g(Yr2005/Lr2005) Labor prod. growth BAU: 3.0 9.5 2.5 2.6 2.5
REF: 2.2 8.4 1.3 2.2 2.0
g(Yr2010/Er2010) Energy prod. growth BAU: 0.9 2.6 0.9 0.8 1.3
REF: 0.8 3.6 0.6 2.2 2.1
Ir2005/Yr2005 Investment to GDP BAU: 20 37 28 29 28
REF: 22 37 23 20 19
RrL2005/Yr2005 Labor inno. expd. BAU: 0.5 1.7 0.8 1.8 3.3
REF: 0.7 0.9 2.2 1.8 2.6
RrE2005/Yr2005 Energy inno. expd. BAU: 0.2 0.6 0.1 0.2 0.2
REF: na na na 0.4 0.5
SrL2005/Yr2005 Labor imit. expd. BAU: 1.2 3.6 0.3 0.2 0.3
SrE2005/Yr2005 Energy imit. expd. BAU: .06 .30 .04 .03 .05
Table 1: Comparison of regional model results for 2005 under BAU with reference valuesREF computed as averages from 1996 to 2006 taken from WDI (2010) and for energy-specific R&D from IEA (2010). g denotes an average yearly growth rate over a five-yearperiod. All values are reported in percent. (We report model results for 2010 in caseof g(Yr2010/Er2010) since the model yields negative energy productivity growth for someregions in 2005 due to initial adjustment effects.)
Europe and in the USA. Furthermore, Figure 5 and Figure 6 in the Appendix illustrate
relevant indicators of the model dynamics. Herein, it is important to note that our
optimization model generates the globally, socially optimal allocation of expenditures
on imitation and innovation. This means, the positive external effect of international
technology spillovers is internalized. International trade, on the contrary, occurs in a
decentralized Nash game.
Basically, the high-income regions USA and EUR follow innovation-based strategies
while the low-income regions INA and CHN follow imitation-based strategies as sug-
gested by Acemoglu et al. (2003, 2006). The reason for this outcome is the advantage
of the high-income countries in terms of education, existing technologies and capital on
the one hand and the advantage of the low-income countries in terms of the potential
to absorb technologies from abroad due to the low quality of their own technologies on
the other hand.
According to Figure 5b, economic growth is in particular substantial in INA. Therein,
14
population growth is a main driver. INA has initially the largest share in global popula-
tion which even increases over time (see Figure 5a). Furthermore, Figure 5c shows our
stylized assumption that INA’s education level catches up to that of the other countries
until 2100. This assumption enhances innovation and imitation via λrt in Equation (8).
Nonetheless, Figure 6a and b show that the productivity growth rates are moderate so
that INA’s productivity stays well below those of the technological leaders USA and
EUR. Herein, labor productivity is equivalent to per capita income which stays low in
INA. Accordingly, population growth is the major driver of the strong long-run GDP
growth.
According to Figure 6c, the USA are clearly the technological leader regarding labor-
specific R&D (innovation) expenditures per GDP, followed by Europe in most periods
(and China in the initial phase of strong economic growth). On the contrary, imitation
plays a minor role in the USA and Europe regarding imitation expenditures per GDP
(Figure 6e/f), while it plays a major role in China and INA. Nonetheless, China also
spends the most on energy-specific innovation per GDP, followed by INA in most periods.
The probable reason is that the energy productivity in these regions is initially very low
so that innovation as well as imitation create a strong benefit per unit invested. The
USA and Europe have initially a rather high energy productivity so that investments in
energy productivity are less beneficial. It is also obvious (Figure 6d/f) that all regions
spend less on both, energy-specific innovation and imitation, as a fraction of GDP over
time, while energy productivity steadily rises over time (Figure 6b).
While data about population, GDP and energy inputs are available across almost
all countries and years under scrutiny, there are limited data on R&D expenditures
in developing countries. Nevertheless, it is well-known that mainly the industrialized
regions drive innovation which is reflected in their R&D expenditures.11 Moreover,
data sources report total R&D expenditures but not labor-specific R&D expenditures as
required for our model. Nevertheless, economic growth is mainly driven by labor-specific
technical progress in our model as in many growth models so that it is a direct substitute
for general technical progress. Energy-specific R&D expenditures are available from
IEA (2010). However, the data cover less than the OECD countries. Finally, there are
probably no data available about expenditures on the adoption and imitation of products
and processes (on a country level). Therefore, we suppose that these expenditures have
a similar magnitude as R&D expenditures. Therein, in our model R&D expenditures
mainly depend on the exponents σR and σS , i.e. a higher exponent creates ceteris paribus
11SEI (2006) reports the global shares in total R&D expenditures of 729 bill. US-$ in the year 2000as follows: North America 39.1, Asia 28.7, Europe 27.9, South America and Caribbean 2.5, Oceania 1.2,Africa 0.6.
15
higher R&D and imitation expenditures. Hence, we reduce σS to 0.01 (compared with
σR = 0.1) so that expenditures on innovation (R&D) and for imitation (adoption) of
technologies per GDP generated by the model have a similar magnitude.
Finally, the strength of technical progress across the factors labor and energy is
adjusted so that it better matches the historical data in terms of labor and energy
productivity growth. Herein, we set λL=1 and λE=3.
3.3 Future scenarios
Table 2 compares our model results with scenarios B1 (B1T1 ASF) and B2 (B2BC
Minicam) by IPCC (2000) which come closest to our scenario among the IPCC scenarios.
Symbol Explanation Scen. Result
L2100 Global population (= labor force) [bill.] BAU: 9.1
B1: 7.1
B2: 10.4
Y2100 Global GDP [trill. US-$] BAU: 300
B1: 471
B2: 354
E2100 Global primary energy cons. p.a. [EJ] BAU: 900
B1: 791
B2: 1370
Q2005−2100 Global cumulated carbon emissions [Gt] BAU: 1258
B1: 1345
B2: 1290
Table 2: Comparison of global model results for 2100 under BAU with reference valuesof scenarios B1 (B1T1 ASF) and B2 (B2BC Minicam) by IPCC (2000).
Scenario B1 assumes low population growth and relatively high economic growth, a
low primary energy intensity, a low carbon intensity and a high fossil fuel availability
in combination with global economic and climate policy solutions. Scenario B2 assumes
medium population growth and medium to low economic growth, a medium to high
primary energy intensity, a balanced carbon intensity and a low fossil fuel availability
in combination with regional economic and climate policy solutions.
In this sense, we follow medium to optimistic assumptions on future socio-economic
developments. The regional time paths of important socio-economic indicators created
by our model are illustrated in Figure 5 in the Appendix. Therein, the primary energy
mix is characterized by the dominance of fossil fuels in (particular coal) for most decades
16
of the century (see Figure 5e in the Appendix). Coal is a cheap and abundant energy
source, and impacts of carbon emissions are not internalized in a business as usual
(BAU) scenario. However, wind and later on even more solar power are significantly
expanded due to learning-by-doing effects. Moreover, biomass is strongly used, mostly
in order to meet the increasing demand on transportation fuels. The resulting baseline
emissions are comparatively low.
4 Assessment
This section applies our model exemplarily to the assessment of (1) a carbon budget-
based climate policy and (2) an analysis od delayed energy-saving technology transfer.
4.1 Carbon budget-based climate policy
We first impose a budget of global emissions cumulated from 2005 to 2100 amounting
to 400Gt of carbon (following Meinshausen et al. 2009 and Allen et al. 2009). The
emissions budget is supposed to translate into a temperature goal of about two degree
in a statistically robust way. Emissions permits are allocated across regions following
a Contraction and Convergence approach (GCI 1990). Therein, per capita emissions
in 2005 follow actually measured per capita emissions in 2000. Per capita emissions
then converge to equal levels across regions until 2050 such that the budget constraint
is fulfilled. We call this climate policy scenario POL.
Herein, the model generates the globally, socially optimal allocation of expenditures
on imitation and innovation. This means, the positive external effect of international
technology spillovers is internalized. Moreover, this globally, socially optimal solution is
independent of distribution matters such as the permit allocation scheme. The permit
allocation scheme of course affects regional consumption losses stemming from climate
policy. Finally, the REMIND model applies the Negishi (1972) algorithm such that re-
gions are not allowed to create debts or surpluses beyond 2100. Therefore, international
transfers can be interpreted as loans that are granted in earlier periods and payed back
in later periods.
Table 3 and Table 4 show the difference between POL as described above and BAU
as discussed in the previous section for relevant indicators. While Table 3 shows the
results for the initial periods, Table 4 shows the results as averages over the time horizon
2005 until 2100. In Table 3, the policy effects have an order of magnitude of around
0.001 to more than one percentage points in terms of growth rates or shares in GDP.
In Table 4, the policy effects have an order of magnitude of around 0.001 to more than
17
Symbol Explanation INA CHN ROW EUR USA
∆g(Yr2005) GDP growth –.02 –.05 –.09 –.12 –.13
∆g(Yr2005/Lr2005) Labor prod. growth –.02 –.05 –.09 –.12 –.13
∆g(Yr2005/Er2005) Energy prod. growth 1.28 1.23 1.12 1.04 .69
∆(Ir2005/Yr2005) Investment to GDP .17 .08 .01 –.08 –.16
∆(RrL2010/Yr2010) Labor inno. expd. –.02 –.03 –.01 –.01 –.03
∆(RrE2005/Yr2005) Energy inno. expd. .02 .02 .01 .01 .05
∆(SrL2010/Yr2010) Labor imit. expd. –.002 –.002 –.003 –.012 –.031
∆(SrE2005/Yr2005) Energy imit. expd. .002 .004 .003 .005 .029
Table 3: Impacts of policy POL (carbon budget) with respect to BAU; changes in yearlygrowth rates and ratios p.a. in the initial years 2005 or 2010 in percentage points. (E.g.a change from 1.100% p.a. to 1.099% p.a. is a –0.001 change in the table. We reportmodel results for 2010 in several cases when the values in 2005 deviate from the generalmodel behavior due to initial adjustment effects.)
0.01 percentage points in terms of growth rates or shares in GDP. Significant changes
can be observed for energy productivity growth which is partly due to shifts in R&D
investments. Obviously, the effects are stronger in earlier periods than in later periods
in accordance with the general behavior of growth models, in which the system initially
changes strongly until a steady state is reached. In both tables, the effects have the
expected signs: Investments in energy-saving innovation and imitation increase due
to the emissions restriction while investments in labor-saving innovation and imitation
decrease.12 As a consequence, GDP growth rates also decrease. Notably, the investment
share in GDP increases in some regions and years (and also on average as shown in Table
4), probably since a higher investment share enhances the implementation of energy-
saving technologies in physical capital as incorporated in our model of technical progress.
We also examine how the directed technical change implemented in our macroeco-
nomic module affects the development of energy conversion technologies in our energy
module. It turns out that technical change directed towards labor productivity leads to
a much stronger and earlier decarbonization of energy supply: In particular the capac-
ities of CCS, furthermore of renewables, expand earlier and more strongly. But labor
productivity growth also results in a higher demand for fossil fuels, in particular in de-
12Nevertheless, there can be single cases with opposite signs in general.
18
Symbol Explanation INA CHN ROW EUR USA
∆∅g(Yr) GDP growth –.003 –.004 –.001 –.002 –.002
∆∅g(Yr/Lr) Labor prod. growth –.003 –.004 –.001 –.002 –.002
∆∅g(Yr/Er) Energy prod. growth .052 .054 –.001 .042 .036
∆∅(Ir/Yr) Investment to GDP .002 .017 .006 .001 .003
∆∅(RrL/Yr) Labor inno. expd. –.006 –.008 –.004 –.008 –.012
∆∅(RrE/Yr) Energy inno. expd. .010 .019 .005 .007 .009
∆∅(SrL/Yr) Labor imit. expd. –.005 –.006 –.001 –.001 –.001
∆∅(SrE/Yr) Energy imit. expd. .004 .007 .002 .001 .002
Table 4: Impacts of policy POL (carbon budget) with respect to BAU; changes in yearlygrowth rates and ratios p.a. are expressed as averages over the time horizon 2005 to2100 in percentage points. (E.g. a change from 1.100% p.a. to 1.099% p.a. is a –0.001change in the table.)
veloping regions (CHN, INA). Technical change directed towards energy productivity
has basically opposite effects: A lower demand for fossil fuels is detrimental for resource
owners (subsumed in ROW). Investments in energy generation capacities (especially
CCS and renewables) are lower (especially under POL) than without energy-specific
technical progress. Overall, all regions benefit from labor- as well as energy- specific
technical progress.
Figure 7 in the Appendix summarizes the most important indicators of BAU shown in
Figures 5 and 6 for the POL scenario. Figure 7a indicates that POL GDP does not visibly
change compared to BAU GDP in Figure 5a. Figure 7b depicts that energy productivity
does significantly rise in POL in all regions. This means, climate policy results in a
reduction of energy intensity via substitution effects and probably to a smaller extent via
technical progress. Figure 7c however shows that the decarbonization of energy supply
is a much stronger effect than the improvement in the energy productivity: Emissions
decrease in all regions to a larger extent than energy productivity increases. Figures 7d
and 7f show that energy-specific innovation and imitation expenditures slightly increase
under POL. (Note the extended scale of the y-axis in these figures compared to 6d and
6f.) Figure 7e finally reveals how the decarbonization is achieved: Biomass, solar and
geothermal expand strongly over time. Wind and hydro, and nuclear (uranium) too,
expand earlier and to a somewhat larger extent. Despite the use of CCS, coal power
19
is tremendously reduced over time. Gas which can serve as a substitute for coal with
lower carbon content is only to a small extent reduced. Oil is reduced to a relatively
small extent too.
Moreover, it has been shown in related previous studies that the introduction of
climate policy results in a re-evaluation of fossil resources. Based on REMIND simula-
tions, this is analyzed in Luken et al. (2011). Changes in the trade of fossil resources
are discussed in Leimbach et al. (2010b). Herein, Luken et al. (2011) find that the
devaluation of tradable fossil energy endowments contributes to a large extent to the
climate policy costs of fossil fuel exporting regions. Moreover, they find that given a
certain emissions target, a reduced availability of low-carbon technologies increases the
volume of emissions permits trade and the magnitude of redistribution effects.
For comparison, we run the following experiment, denoted by ’Fixed’:13 We run
BAU. Then we fix the expenditures on innovation and imitation (Rrit, Srit) at their
BAU levels and run POL. We compute differences between POL and BAU and compare
the results with the differences between POL and BAU computed previously, denoted
by ’Endogenous’. It turns out that all regions benefit from policy induced technical
progress in experiment ’Endogenous’ compared to ’Fixed’. However, the difference in
mitigation costs between ’Fixed’ and ’Endogenous’ has a similarly small magnitude as
in the results presented in Table 3 and Table 4, i.e. about 0.01 percentage points.
Finally, in an additional experiment, we take the investment over capital ( IK ) ratio
for each year from a benchmark run and fix it. Hence, it appears as a time variant
parameter in the equation that governs technical progress. (In the usual run, IK is
endogenous.) Then we run a climate policy scenario. Based on the numerical results
and on theoretical intuition, we identify the following effects:
Since capital investment I does not only contribute to the capital stock K but also
to the knowledge (technology) stock A, each marginal unit of I creates an additional
marginal beyond the pure expansion of the capital stock. This applies in particular to
earlier periods, because rising A (which does not depreciate like K) creates a benefit
in all future periods. Thus, comparing our model where capital investment enhances
technical progress to a usual Ramsey model with endogenous capital investment, we
find: I will on average increase, and it will be shifted from future to present periods.
Climate policy can enhance this effect, in particular with respect to energy productivity.
Additionally, climate policy leads to higher investments in capacities of low-carbon tech-
nologies. These investments rival the investment in energy-specific technical progress.
This effect counteracts the effect described above. However, the effects are relatively
13Not explicitly shown in the tables.
20
small in numerical terms. Independent of such policy-induced effects, the inclusion of
physical capital in the function that describes technical progress significantly influences
the baseline (BAU) path of technical progress. Herein, Figure 5d shows that all coun-
tries except INA have initially higher investment ratios ( IY and thus I
K ) than in the
long-run steady state. The rise in IK creates an additional growth enhancing effect in
early periods and enhances the ’steady state convergence’ behavior of the model.
4.2 Delayed energy-saving technology transfer
In the following, we will assess in how far postponing energy-specific international tech-
nology transfer affects mitigation costs given the previous climate policy. International
technology transfer requires a sufficient absorptive capacity determined among other
factors by the business and legal environment. It requires investments in the absorption
and imitation of technologies and in physical capital financed from national or interna-
tional sources. Herein, climate policy can play an active role in creating the necessary
absorptive capacity and fostering national and international financing.
Accordingly, in our model, we interpret the spillover term(AtAt
)σA
as the channel
of international technology transfer under scrutiny. Energy-specific imitation expendi-
tures SrEt enable the use of this channel and can be financed within each country as
well as through international transfers (in form of the composite commodity) within
the general budget (Equation 9). Postponing international technology transfer is rep-
resented in the following stylized way: Energy-specific imitation expenditures SrEt are
exogenously bound to a value close to zero for all regions and all periods before each
year t0 ∈ {2010, 2015, 2020, 2025, 2030, 2035, 2040}. This means, we run one additional
BAU and POL scenario for each of the seven delay periods. From t0 on, energy-specific
interregional technology diffusion evolves endogenously as before. Herein, the relaxation
of imitation expenditures at t0 is anticipated.
Figure 1 plots the regionally different effects of postponing energy-specific technology
diffusion in form of the difference between consumption in a baseline scenario BAU
where financing is postponed versus consumption in the baseline scenario BAU where
financing starts immediately in 2005 relative to consumption in the latter scenario. In
all calculations of consumption losses, we cumulate consumption losses between 2005
and 2100 and discount at a rate of 3% per year.
Obviously, postponing creates consumption losses that range from less than 0.1 to
more than 0.5 percentage points for all regions due to a higher energy demand per out-
put since energy-specific technical progress is hindered. Accordingly, early investment
in energy-saving technology diffusion is beneficial for all regions, given our model setup.
21
0 0.1 0.2 0.3 0.4 0.5 0.6
World
ROW
INA
CHN
EUR
USA
Consumption loss [%]
20052010201520202025203020352040
Figure 1: Regional effects of postponing international transfers of energy efficienttechnologies. Consumption losses are reported as the percentage change betweenconsumption in a baseline scenario BAU where transfers are postponed (as indicatedin the legend) and consumption in the baseline scenario BAU where transfers startimmediately in 2005. Consumption losses are cumulated from 2005 to 2100 anddiscounted at a rate of 3% p.a.
This is probably due to the following reasons: First, our model setup allows all regions
to benefit from the global knowledge pool, in this case regarding energy-specific techno-
logical knowledge. Second, regions can benefit from technical progress in other regions
through international transfers or in other words through commodity trade.
As expected, China suffers most from postponing international technology diffusion
because it starts at a low energy productivity and is able to catch up fast, followed by the
developing region INA. Europe suffers least due to its good initial energy productivity,
followed by the USA and the Rest of the World. The gains from financing technology
transfer appear to be higher in earlier periods since the process of growth and technolog-
ical catching up is more pronounced in earlier periods than in later periods. Intuitively,
early investments in technical progress are beneficial over the whole time horizon while
late investments have a limited remaining scope.
Figure 2 plots the regional effects of climate policy POL as the difference between
POL and BAU consumption relative to BAU consumption for each start date of
financing. Consumption losses obviously slightly rise when postponing the financing of
international technology diffusion. Basically, Figure 2 illustrates that our integrated
assessment model generates consumption losses of less than one percent for all regions
22
0 0.2 0.4 0.6 0.8 1 1.2
World
ROW
INA
CHN
EUR
USA
Consumption loss [%]
20052010201520202025203020352040
Figure 2: Regional effects of climate policy POL (carbon budget) for different start dates(as indicated in the legend) of international transfers of energy efficient technologies.Consumption losses are reported as the percentage change between POL and BAUconsumption for each start date of transfers.
except China. China is accordingly affected most severely, followed by the region
Rest of the World and the USA. Europe can probably benefit from its good energy
productivity and the developing region INA from its low per-capita emissions in the
context of interregional permit trading so that mitigation costs are low. However, most
of the consumption gains from early technology transfer have already been exploited in
BAU.
5 Sensitivity
This section first critically discusses model characteristics and then addresses them in a
sensitivity analysis.
5.1 Critical discussion
Our integrated assessment model has the following characteristics:
23
We represent endogenous innovation and imitation and thus international technology
transfer in a stylized fashion. Our functional forms of modeling innovation and imitation
follow the Schumpeterian model of endogenous growth. Other functional forms might
lead to a different dynamic behavior, though. Also, from a micro-economic point of view,
the implementation of the Schumpeterian model does not take profit maximization of
firms and monopolistic power in intermediate production explicitly into account. In
fact, our model reproduces the typical behavior of endogenous growth models on the
macro level.
Moreover, in each region production is specified in form of a standard CES structure
(as illustrated in Figure 4 in the Appendix) that assumes certain elasticities of substi-
tution which determine the possibility of switching between energy, capital and labor
and between specific energy inputs. This structure and the elasticity values influence
mitigation costs.
In general, the calibration of any complex, dynamic, numerical model involves un-
certainties. This is especially true with respect to modeling innovation and international
technology diffusion and related expenditures. We build our calibration on econometric
estimates. These estimates provide a range of appropriate parameter values but the
estimated models do not match our model one by one. Furthermore, the comparison of
model outcomes with reference data reveals a good match in most cases. Initial growth
rates of energy productivity in Europe and the USA are slightly lower than in the refer-
ence data. In general, the sensitivity of the model with respect to policy induced effects
appears small.
Basically, our model shows the typical behavior of North-South growth models, i.e.
strong adjustment processes in early periods in terms of investment in capital and tech-
nology, and strong North-South transfers (from USA, EUR and ROW to INA and CHN;
c.f. Lucas 1990). Consequently, most of the policy induced technology effects occur in
early periods, too, while long-run growth paths evolve at low growth rates (c.f. Figure
6 in the Appendix) and are hardly affected by the policy experiments. On the con-
trary, under the assumption of distant future economic growth at a constant high rate,
GDP and emissions and the resulting mitigation costs would be higher (c.f. Hubler
2011). Moreover, discounting gives a lower weight to future consumption and its pol-
icy induced changes and consequently influences mitigation costs as in every economic
long-run analysis.
Furthermore, we only capture expenditures on the innovation and imitation of tech-
nologies that improve energy productivity on a macro-economic scale. We leave aside
the direct international transfer of energy technologies like wind and solar power in
24
this analysis. Taking this into account, would presumably strongly increase policy in-
duced innovation and imitation and could further reduce mitigation costs. One would
in particular expect a shift of innovation from fossil to non-fossil technologies. The
REMIND model encompasses a full energy system module, though, which enables the
early expansion of renewable energies and the decarbonization of energy supply in every
region. Herein, the energy system module is calibrated to benchmark data. The future
development of the energy mix represents only one exemplary scenario though.
Our policy analysis suggests relatively low mitigation costs of keeping a carbon
budget of 400Gt for the time period 2005 to 2100, given a baseline scenario of relatively
low emissions.
Therefore, in summary all policy results need to be taken with some caution. They
need to be interpreted with respect to the baseline scenario that we have calibrated
based on econometric, historical and scenario data.
5.2 Sensitivity analysis
In the following, we address the aspects discussed above by carrying out a detailed
sensitivity analysis for regional consumption losses stemming from climate policy POL
based on the BAU scenario ’Default’ that we have examined so far:
(1) We switch off the availability of all renewable energies and CCS (of coal, gas and
biomass) in all regions (’-Renew’). (2) We change the constant elasticity of substitution
in the upper CES level (c.f. Equation 7 and Figure 4 in the Appendix) to 0.2 and
alternatively to 0.8, while our standard value is 0.5. (3) We vary the elasticity of
technical progress with respect to related investments governed by the exponents σR
and σS simultaneously by the same factors, namely two (twice the previous value) and
0.75. (4) We vary the strength of energy- and labor-specific innovation governed by
λR and (5) the strength of energy- and labor-specific imitation governed by λS by the
factors 1.5 and 0.75. (6) Finally, we raise the strength of energy-specific innovation as
well as imitation – keeping the strength of labor-specific technical progress unchanged
– by the factor 1.5. We then reduce it to one third so that it has the same strength as
for labor (λE = λL). Herein, the range of the parameter value variations is limited by
the capability of finding feasible and optimal solutions for the optimization problem as
well as by economic reasoning.
The results are reported in Table 5 which shows consumption losses between POL
and BAU, cumulated from 2005 until 2100 and discounted at a rate of 3% per year.14
14Energy-specific technology transfer is not delayed but allowed from 2005 on. Therefore, scenario
25
(1) (2) (3)Region Default -Renew σY = 0.2 σY = 0.8 2 · σR/S 0.75 · σR/S
USA 0.69 1.93 0.45 1.20 0.77 0.67EUR 0.37 1.32 0.28 0.52 0.40 0.36CHN 1.13 4.14 0.49 2.48 1.25 1.09INA 0.39 1.55 0.28 0.46 0.42 0.38ROW 0.99 0.29 0.30 2.11 1.08 0.95World 0.65 1.64 0.34 1.20 0.71 0.63
(4) (5) (6)Region 1.5 · λR 0.75 · λR 1.5 · λS 0.75 · λS 1.5 · λE 0.33 · λE
USA 0.65 0.71 0.63 0.72 0.48 1.20EUR 0.36 0.38 0.35 0.39 0.28 0.53CHN 1.07 1.15 1.05 1.15 0.76 2.05INA 0.39 0.39 0.44 0.32 0.27 0.68ROW 0.94 1.01 0.88 1.01 0.66 1.64World 0.62 0.66 0.62 0.65 0.45 1.10
Table 5: Sensitivity analysis for regional mitigation costs reported as the percentagechange between POL and BAU consumption. The losses are cumulated from 2005 to2100 and discounted at a rate of 3% p.a.
The baseline scenario ’Default’ is the same as in the previous analysis. Accordingly,
global consumption losses exceed 0.5%. The most striking increase in mitigation costs
to more than 1.5% of global consumption occurs when switching of the availability of
renewable energies and CCS in experiment (1) ’-Renew’. Herein, China’s consumption
losses even rise to more than 4%. Moreover, a significant increase in global consumption
losses to more than 1% occurs when raising the elasticity of substitution in the upper
CES nest in (2). Herein, the peak of global carbon emissions in BAU rises from below
16Gt in the standard scenario to almost 20Gt in experiment (2) where the elasticity
of substitution is raised (σY = 0.8). Obviously, the higher flexibility in the production
structure already increases BAU production and emissions which overcompensates the
resulting improved possibility to replace fossil fuel inputs in POL. The opposite applies
to the scenario where the elasticity of substitution is reduced (σY = 0.2).
On the contrary, the impact of the variation in most of the coefficients and exponents
within our model of endogenous growth is surprisingly small. An exemption is the
strength of energy-specific technical progress (based on innovation as well as imitation)
’Default’ resembles the 2005 result in Figure 2.
26
for a given strength of labor-specific technical progress as examined in experiment (6).
Accordingly, setting the coefficient that governs the strength of energy-specific technical
progress equivalent to the coefficient of labor-specific technical progress (0.33 ·λE = λL)
raises global losses to more than 1% and China’s losses to about 2%. On the contrary,
mitigation costs are hardly affected in experiments (4) and (5) where the strength of
energy and labor-specific innovation and imitation is varied simultaneously to the same
extent. This result emphasizes the role of energy-specific technical progress relative to
general or labor-specific technical progress with respect to mitigation costs.
6 Conclusion
We introduce endogenous directed technical change into integrated assessment based
on theory and evidence. We distinguish innovation and imitation, in other words in-
ternational technology transfer, and take the role of capital investment in creating and
implementing new technologies into account. We study the regional pattern of miti-
gation costs in form of consumption losses induced by a global carbon budget-based
climate policy.
Mitigation costs turn out to be robust with respect to a variation in the parameter
values within our model of endogenous growth – except the coefficient of energy-specific
relative to the coefficient of labor-specific innovation and imitation expenditures. This
result suggests that a higher effectivity (in terms of technical progress per unit of in-
vestment) of energy-specific innovation and imitation expenditures can overcompensate
rising emissions due to economic growth and thus reduce mitigation costs. However,
the effectivity is exogenously determined by technological restrictions and at best partly
susceptible by economic policy in the short-run in reality. In the medium- to long-run,
the effectivity could be improved by emphasizing energy efficiency aspects in educa-
tion, basic research, infrastructure, structural change and capacity building for R&D
and for the absorption of foreign technologies. While our analysis exogenously adjusts
a regional parameter representing education and absorptive capacity, further research
may endogenize such a parameter in order to represent fundamental aspects of long-run
development.
Motivated by the announcement of providing financial and technological transfers
for developing countries at the Cancun 2010 summit, we examine interregional energy-
saving technology transfer starting at different points of time. Herein, in general, all
regions gain from technology transfer due to its growth effect. China appears as the main
beneficiary of early technology transfer, followed by the region of the developing coun-
27
tries. These results suggest that enabling energy-saving technology transfer as soon as
possible supports developing countries. The decarbonization of economic development,
however, additionally requires the switch from fossil energies to renewable energies. In
our model setup, such renewable energies are – to regionally different extents – available
in all regions. Further research may therefore assess in how far the international trans-
fer of renewable energies and the timing of such transfers can affect regional mitigation
costs.
In general, our results suggest that endogenous energy-specific technical progress
has already been strongly exploited in the business as usual (BAU) baseline due to its
positive effect on consumption. This might be due to perfect foresight and regionally
weighted global optimization in combination with energy scarcity. In reality, these
baseline gains are not fully exploited on a market base due to risks and frictions
and missing information (for possible baseline gains see McKinsey&Company 2009).
Realizing these gains may require active policy support in form of an integrated
climate and energy policy concept. Due to these baseline gains, the early support of
energy-saving technology transfer could be a ’carrot’ to encourage developing countries
to engage in climate protection.
7 References
Acemoglu, D., P. Aghion and F. Zilibotti (2003). Vertical Integration and Distance toFrontier. Journal of the European Economic Association 1(2-3), 630-638.
Acemoglu, D., P. Aghion and F. Zilibotti, (2006). Distance to Frontier, Selection, andEconomic Growth. Journal of the European Economic Association 4(1), 37-74.
Acemoglu, D. (2002). Directed Technical Change. The Review of Economic Studies69(4), 781-809.
Acemoglu, D. (2009). Introduction to Modern Economic Growth. Princeton UniversityPress, New Jersey, USA.
Aghion, P. and P. Howitt (1992). A Model of Growth through Creative Destruction.Econometrica 60(2), 323-351.
Aghion, P. and P. Howitt (2009). The Economics of Growth. The MIT Press,Cambridge, Massachusetts, USA and London, England.
Allen, M.R., D.J. Frame, C. Huntingford, C.D. Jones, J.A. Lowe, M. Meinshausen andN. Meinshausen (2009). Warming caused by cumulative carbon emissions towards thetrillionth tonne. Nature 458(7242), 1163.
Arrow, K. (1962). The Economic Implications of Learning by Doing. Review of
28
Economic Studies 28, 155-173.
Barreto, L. (2001). Technological Learning in Energy Optimisation Models andDeployment of Emerging Technologies. PhD Thesis ETH Zurich 14151, 277 pp.,http://dx.doi.org/10.3929/ethz-a-004215893.
Bauer, N., O. Edenhofer S. and Kypreos (2006). Linking Energy System and Macroeco-nomic Growth Models. Journal of Computational Management Science (Special Issueon Managing Energy and the Environment) 5(1-2), 95-117.
Benhabib, J. and M. Spiegel (2005). Human capital and technology diffusion. In: P.Aghion and S. Durlauf (eds.), Handbook of Economic Growth, Elsevier, chapter 13.
Bosetti, V., C. Carraro, M. Galeotti, E. Massetti and M. Tavoni (2006). WITCH: AWorld Induced Technical Change Hybrid Model. The Energy Journal, Hybrid Modelingof Energy-Environment Policies: Reconciling Bottom-up and Top-Down, 13-38.
Bosetti, V., C. Carraro, E. Massetti and M. Tavoni (2008). International energy R&Dspillovers and the economics of greenhouse gas atmospheric stabilization. EnergyEconomics 30(6), 2912-2929.
Bosetti, V., C. Carraro, R. Duval and M. Tavoni (2011). What should we expectfrom innovation? A model-based assessment of the environmental and mitigation costimplications of climate-related R&D. Energy Economics 33(6), 1313-1320.
Chichilnisky, G. and G. Heal (eds.) (2000). Environmental Markets - Equity andEfficiency. Columbia University Press, New York, USA.
Chichilnisky, G. and K.A. Sheeran (2009). Saving Kyoto: An Insider’s Guide to How ItWorks, Why It Matters and What It Means for the Future. New Holland Publishers,London, Cape Town, Sydney, Auckland.
Crassous, R., J.-C. Hourcade and O. Sassi (2006). Endogenous Structural Changeand Climate Change Targets – Modeling Experiments with Imaclim-R. The EnergyJournal 27, Special Issue: Endogenous Technological Change and the Economics ofAthmospheric Stabilisation, 259-276.
Crespo J., C. Martin and F. Velazquez, (2004). International technology spillovers fromtrade: the importance of the technological gap. Investigaciones Economicas, FundacionSEPI 28(3), 515-533.
Diao, X., J. Rattsø and H.E. Stokke (2005). International spillovers, productivitygrowth and openness in Thailand: an intertemporal general equilibrium analysis.Journal of Development Economics 76, 429-450.
Edenhofer, O., N. Bauer and E. Kriegler (2005). The impact of technological change onclimate protection and welfare: Insights from the model MIND. Ecological Economics54, 277-292.
Edenhofer, O., C. Carraro, J.-C. Hourcade, K. Neuhoff, G. Luderer, C. Flachsland,M. Jakob, A. Popp, J. Steckel, J. Strohschein, N. Bauer, S. Brunner, M. Leimbach,H. Lotze-Campen, V. Bosetti, E. de Cian, M. Tavoni, O. Sassi, H. Waisman, R.Crassous-Doerfler, S. Monjon, S. Droge, H. van Essen, P. del Rio, A. Turk (2009).RECIPE - The Economics of Decarbonization, Synthesis Report, 96 pp.
29
Gerlagh, R. (2008). A climate-change policy induced shift from innovations in carbon-energy production to carbon-energy savings. Energy Economics 30, 425-448.
GCI (1990). Contraction and Convergence (C&C) is the science-based, global climatepolicy framework proposed to the UN since 1990 by the Global Commons Institute.http://www.gci.org.uk/.
Goulder, L.H. and S.H. Schneider (1999). Induced Technological Change and theAttractiveness of CO2 Abatement Policies. Resource and Energy Economics 21,211-253.
Greiner, A. and W. Semmler (2002). Externalities of investment, education andeconomic growth. Economic Modelling 19, 709-724.
Griffith, R., S. Redding and J. van Reenen (2003). R&D and Absorptive Capacity:Theory and Empirical Evidence. Scandinavian Journal of Economics 105(1), 99-118.
Griffith, R., S. Redding and J. van Reenen (2004). Mapping the Two Faces of R&D:Productivity Growth in a Panel of OECD Industries. The Review of Economics andStatistics 86(4), 883-895.
Griliches, Z. and F. Lichtenberg (1984). Interindustry Technology Flows and Produc-tivity Growth: A Reexamination. Review of Economics and Statistics 66(2), 324-329.
Grossman, G. and E. Helpman (1991). Innovation and Growth in the Global Economy.The MIT Press, Cambridge, Massachusetts, USA and London, England.
Hubler, M. (2011). Technology Diffusion under Contraction and Convergence: A CGEAnalysis of China. Energy Economics 33(1), 131-142.
IEA (2010). Data Services. International Energy Agency, Paris, France. http://data.iea.org/ieastore/default.asp.
IPCC (2000). Special Report on Emissions Scenarios (SRES). Intergovernmental Panelon Climate Change, Geneva, Switzerland.http://www.grida.no/publications/other/ipcc_sr/?src=/climate/ipcc/
emission/.
Kahouli-Brahmi, S. (2008). Technological learning in energy-environment-economymodelling: A survey. Energy Policy 36, 138-162.
Kemfert, C. (2005). Induced technological change in a multi-regional, multi-sectoral,integrated assessment model (WIAGEM) Impact assessment of climate policy strate-gies. Ecological Economics 54, 293-305.
Kneller, M. (2005). Frontier technology, Absorptive Capacity and Distance, OxfordBulletin of Economics and Statistics 67(1), 1-23.
Kossoy, A. and P. Ambrosi (2010). State and Trends of the Carbon Market 2010.Carbon Finance at the World Bank, Environment Department, Washington, DC, USA.
Leimbach, M. and Baumstark, L. (2010). The impact of capital trade and technologicalspillovers on climate policies. Ecological Economics 69, 2341-2355.
30
Leimbach, M., N. Bauer, L. Baumstark and O. Edenhofer (2010a). Mitigation Costsin a Globalizing World: Climate Policy Analysis with REMIND-R. EnvironmentalModeling and Assessment 15, 155-173.
Leimbach, M., N. Bauer, L. Baumstark, M. Luken and O. Edenhofer (2010b). Tech-nological Change and International Trade: Insights from REMIND-R. The EnergyJournal 31, Special Issue: The Economics of Low Stabilization, 109-136.
Loschel, A., and V.M. Otto (2009). Technological Uncertainty and Cost-Effectivenessof CO2 Emission Reduction. Energy Economics 31, 4-17.
Lucas, R. (1990). Why Doesn’t Capital Flow from Rich to Poor Countries? TheAmerican Economic Review 80(2), 92-96.
Luken, M., O. Edenhofer, B. Knopf, M. Leimbach, G. Luderer and N. Bauer (2011).The role of technological availability for the distributive impacts on climate changemitigation policy. Energy Policy 39, 6030-6039.
McKinsey&Company (2009). Pathways to a Low-Carbon Economy. Version 2 of theGlobal Greenhouse Gas Abatement Cost Curve, 192 pp.https://solutions.mckinsey.com/climatedesk/default.aspx.
Meinshausen, M., N. Meinshausen, W., Hare, S.C.B. Raper, K. Frieler, R. Knutti, D.J.Frame, M.R. Allen (2009). Greenhouse-gas emission targets for limiting global warmingto 2oC. Nature 458, 1158-1163.
Negishi, T. (1972). General Equilibrium Theory and International Trade. North-Holland Publishing Company, Amsterdam.
Nelson, R. and E. Phelps (1966). Investment in Humans, Technological Diffusion, andEconomic Growth. The American Economic Review, Papers and Proceedings 61, 69-75.
Nicita, L., C. Carraro and E. Massetti (2009). How Does Climate Policy AffectTechnical Change? An Analysis of the Direction and Pace of Technical Progress in aClimate-Economy Model. The Energy Journal, Climate Change Policies after 2012,7-38.
Otto, V.M., Dellink, R. and A. Loschel (2007). Energy Biased Technical Change - ACGE Analysis. Resource and Energy Economics 29 (2), 137-158.
Otto, V.M., A. Loschel and J. Reilly (2008). Directed Technical Change and Differen-tiated Climate Policy. Energy Economics 30 (6), 2855-2878.
Popp, D. (2004). ENTICE: endogenous technological change in the DICE model ofglobal warming. Journal of Environmental Economics and Management 48, 742-768.
Popp, D. (2006). Innovation in climate policy models: Implementing lessons from theeconomics of R&D. Energy Economics 28, 596-609.
Romer, P. (1990). Endogenous Technological Change. Journal of Political Economy 98,71-102.
Schankerman, M. (1981), The Effects of Double-counting and Expensing on the
31
Measured Returns to R&D. Review of Economics and Statistics 63(3), 454-458.
Scherer, F. (1982). Inter-industry Technology Flows and Productivity Growth. Reviewof Economics and Statistics 64(4), 627-634.
Scherer, F. (1984). Using Linked Patent and R&D Data to Measure Inter-industryTechnology Flows. In: Z. Griliches (ed.), R&D, Patents, and Productivity, NBER andUniversity of Chicago Press, Chicago, USA.
SEI (2006). Science and Engineering Indicators. National Science Foundation, USA.http://www.nsf.gov/statistics/seind06/c4/c4s6.htm.
Tanaka, K. and E. Kriegler (2007). Aggregated Carbon Cycle, Atmospheric Chemistry,and Climate Model (ACC2). Reports on Earth System Science, 40. Max-Planck-Institute of Meteorology, Hamburg, Germany.
UN (2008). World Population Prospects - The 2008 Revision. United Nations, NewYork, USA. http://www.un.org/esa/population/publications/wpp2008/wpp2008_highlights.pdf.
WDI (2010). World Development Indicators. The World Bank, Washington DC, USA.http://data.worldbank.org/data-catalog/world-development-indicators.
World Bank (2010). Climate Investment Funds (CIF). The World Bank, WashingtonDC, USA. http : //www.climateinvestmentfunds.org/cif/node/2.
Zachariadis, M. (2003). R&D, Innovation, and Technological Progress: A Test of theSchumpeterian Framework without Scale Effects. The Canadian Journal of Economics36(3), 566-586.
32
8 Appendix
Symbol Explanation
t = {2005, 2010, ..2100} Time (years in 5-year steps)
r = {INA,CHN, Regions (INA: Africa, Latin America,
ROW,EUR,USA} India and other Asia - CHN: China -
ROW: Middle East, Japan, Russia, rest of the world -
EUR: Europe - USA: United States of America)
i = {L,E} Factors affected by technical progress (labor, energy)
Yrt Production (income)
Mrt Net imports
Crt Consumption
Arit Technology level
Ait Average global technology level (technology pool)
Krt Capital input
Lrt Labor input
Ert Energy input
Irt Investment in capital
Qrt Investment in the energy system
Rrit Innovation or R&D expenditures
Srit Imitation expenditures
Table 6: Sets and endogenous variables.
33
Symbol Explanation Value
λi Factor specialty of technical progress L: 1 E: 3
λr2005 Education level in 2005 INA: 0.3 CHN: 0.7
ROW: 0.4 EUR: 0.75
USA: 0.9
λR/λS Coefficient of innovation/imitation expd. R: 0.4 S: 0.12
σR/σS Exponent of innovation/imitation expd. R: 0.1 S: 0.01
σA/σI Exponent of tech. gap/investment A: 1 I: 1
Table 7: Exogenous parameters.
34
Figure 3: The REMIND modules and their interaction.
35
Figure 4: The CES structure of the regional REMIND production function; σ indicatesconstant elasticities of substitution.
36
2020 2040 2060 2080 21000
2
4
6
8
10
Year
Pop
ulat
ion
[Bill
.]
ROWCHNINAEURUSA
(a)
2020 2040 2060 2080 21000
20
40
60
80
100
120
140
YearG
DP
[tr
ill.
US
−$
]
EURUSAROWINACHN
(b)
2020 2040 2060 2080 21000
0.2
0.4
0.6
0.8
1
Year
Educa
tion le
vel
EURUSAROWINACHN
(c)
2020 2040 2060 2080 21000
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Year
Inve
stm
en
t p
er
GD
P
EURUSAROWINACHN
(d)
2020 2040 2060 2080 21000
200
400
600
800
1000
Year
Prim
ary
ener
gy c
onsu
mpt
ion
[EJ]
BiomassSolarWindHydroGeotherm.UraniumCoalNat. GasOil
(e)
2020 2040 2060 2080 21000
2
4
6
8
10
12
14
16
Year
Car
bon
Em
issi
ons
[Gt]
ROWCHNINAEURUSA
(f)
Figure 5: Simulation results for BAU.
37
2020 2040 2060 2080 21000
20
40
60
80
100
YearLabo
r pr
od. (
per
capi
ta in
com
e) [t
sd. U
S−
$]
EURUSAROWINACHN
(a)
2020 2040 2060 2080 21000
10
20
30
40
50
60
70
80
YearE
nerg
y pr
oduc
tivity
[US
−$
/ MJ]
EURUSAROWINACHN
(b)
2020 2040 2060 2080 21000.005
0.01
0.015
0.02
0.025
0.03
0.035
Year
La
bo
r R
&D
exp
en
ditu
res
per
GD
P
EURUSAROWINACHN
(c)
2020 2040 2060 2080 21000
1
2
3
4
5
6x 10
−3
Year
Ene
rgy
R&
D e
xpen
ditu
res
per
GD
P
EURUSAROWINACHN
(d)
2020 2040 2060 2080 21000
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
Year
La
bo
r te
ch im
itatio
n e
xpen
d p
er
GD
P
EURUSAROWINACHN
(e)
2020 2040 2060 2080 21000
0.5
1
1.5
2
2.5
3
3.5x 10
−3
Year
Ene
rgy
tech
imita
tion
expe
nd p
er G
DP
EURUSAROWINACHN
(f)
Figure 6: Simulation results for BAU.
38
2020 2040 2060 2080 21000
20
40
60
80
100
120
140
Year
GD
P [
trill
. U
S−
$]
EURUSAROWINACHN
(a)
2020 2040 2060 2080 21000
10
20
30
40
50
60
70
80
YearE
nerg
y pr
oduc
tivity
[US
−$
/ MJ]
EURUSAROWINACHN
(b)
2020 2040 2060 2080 2100−2
0
2
4
6
8
Year
Car
bon
Em
issi
ons
[Gt]
ROWCHNINAEURUSA
(c)
2020 2040 2060 2080 21000
1
2
3
4
5
6
7x 10
−3
Year
Ene
rgy
R&
D e
xpen
ditu
res
per
GD
P
EURUSAROWINACHN
(d)
2020 2040 2060 2080 21000
100
200
300
400
500
600
700
800
Year
Prim
ary
ener
gy c
onsu
mpt
ion
[EJ]
BiomassSolarWindHydroGeotherm.UraniumCoalNat. GasOil
(e)
2020 2040 2060 2080 21000
0.5
1
1.5
2
2.5
3
3.5
4x 10
−3
Year
Ene
rgy
tech
imita
tion
expe
nd p
er G
DP
EURUSAROWINACHN
(f)
Figure 7: Simulation results for POL.
39
Top Related